Because the odds are so low that a rejected contender will win a wildcard spot, it removes any unconscious scoring bias you might implement. In other words, I was thinking of the scenario in which you might only make some of the rejects eligible for a wildcard spot based on a scoresheet. You might unconsciously bias your scoring to make some rejects eligible and others not. The beauty of the wildcard is that it's totally random.
EDIT: Another thought. Wildcard entries would set a nice benchmark for merit-based entrants to YC to compete against. They would have more to prove because they've been written off. I'd be curious how this would affect the larger group dynamics of the programme.
Even if this were true, which I wouldn't argue against, it doesn't mean that it's where the mistakes of the highest aggregate value occur.
For example, what percentage of first phase rejections try again, as compared with second phase rejections?
But the trouble with that is that, unlike funding, interview slots are constrained.
You could apply the usual interview scaling technique of adding another phase (e.g. shorter pre-screen with a single interviewer) or delegating. This has the obvious disadvantages, which is, arguably, one of the major reasons big company hiring processes are so unreliable.